By conventional assessments, in vitro fertilization (IVF) is one of the safest medical treatments. But producing new life brings immense responsibilities. Recently, there have been disquieting reports of foetal abnormality after these treatments and here we evaluate the potential risks associated with intracytoplasmic sperm injection (ICSI), embryo freezing and pre-implantation genetic diagnosis. In our opinion, before translating new techniques into practice, more research, particularly in animals, is desirable. In addition, better child follow-up and a fresh approach to regulation are also needed.

The latest publication highlighting the potential risks of IVF and ICSI¹ should cause serious unease. Hansen et al. report that 75 out of 837 infants born after IVF (9%) and 26 out of 301 infants born after ICSI (8.6%) had a major birth defect; a control group of 4000 naturally conceived foetuses from a similar population (the figures included medically terminated pregnancies) had half the chance of a major defect (4.5%). Reports of a treatment that leaves almost 10% of children handicapped in some way must not be dismissed as a mere statistical quirk. And what's more, this disquieting trend was still apparent when these worrying figures were corrected for multiple births, maternal age and parity � even IVF singletons were at considerable risk.

Earlier studies had been more reassuring. The original Medical Research Council study², which catalogued all UK births between 1978 and 1987 for the first year of life, was very encouraging � 2.2% of 1581 births showed a major congenital malformation, which was no more than population-based estimates of the prevalence in the population at large. Very similar data were also published from studies in France³ (2.8%), Australia⁴ (2.2% malformations in the first 2242 births after IVF in that country), and Israel⁵ (2.2% of 1475 babies born after IVF). All these countries have good health care and sound medical follow-up, so there was every reason to consider IVF as essentially safe for a singleton foetus. However, early studies showed that even singleton IVF pregnancies had an increased risk of preterm delivery, with more of the babies being small for their gestational age and of lower birth-weight than babies delivered after natural conception^6,7.

Some recent reports have, however, raised more concerns. A study of 5856 babies in Sweden showed that 5.4% of babies born after IVF had a serious malformation and that the incidence of neural tube defect and oesophageal atresia was higher than in a control group⁸. Because one third of these babies were born prematurely and 27% of the births were from a multiple conception, the feeling was that these adverse findings were probably a result of the usual practice of transferring several embryos. Similarly, the most recent study, which identified a high malformation rate (7.3%) after ICSI⁸, was not as alarming because it seemed that both the rather high perinatal mortality rate they reported and the malformation rate were probably associated with the high multiple birth rate they recorded. The message seemed clear � reduce the number of embryos transferred after IVF and the morbidity associated with assisted reproduction would be no higher than that after natural conception.

But this most recent study cannot be dismissed so easily. And it is worrying that it has attracted so little attention from those practising and regulating IVF, compared with the scrutiny given it by the lay press and those ready to criticise this challenging technology. One problem is the nature of infertility medicine. Patients who are infertile express extraordinary desperation. The inability to have a child of one's own challenges the most basic instinct � the need to pass on one's genes � and many couples are prepared to try almost anything, no matter how speculative, unproven or even risky the technology. Medical practitioners, often out of genuine sympathy, are sometimes too ready to avoid crushing hope and instead offer treatments with little or no evidence base. It also has to be said that many people suspect, rightly or wrongly, that because almost universally, infertility medicine is not fully funded publicly, there are commercial pressures driving many treatments. Moreover, some therapies, which on any assessment are highly empirical or unproven, are being used in human subjects before they have been validated by proper cell culture experiments or detailed animal research.

One of the questions that must be asked about the data from Hansen et al.¹ is whether there is anything that is specific to Western Australian practice that could have given rise to such a high anomaly rate. This seems unlikely, given that these figures were derived from three IVF clinics practising conventional IVF techniques (Figs 1,2). The report also lacks some important information. For example, it does not record the peak oestradiol levels in these IVF patients during the treatment, the amount of the exogenous chorionic gonadotrophin stimulus, the method of embryo culture or how many of these patients underwent embryo replacement after cryopreservation. These are among several factors that might increase the potential risk to a subsequent pregnancy and which need to be assessed.

Fig. 1
		Figure 1 | Egg retrieval, IVF, embryo culture and transfer. Embryos are generally transferred to the uterus either 2 or 5 days after egg retrieval.

Fig. 2
		Figure 2 | Techniques used in assisted reproduction. IVF involves the incubation of egg with sperm, allowing fertilization to occur naturally. For men with low numbers of sperm, eggs can be fertilized through the ICSI technique, where a single sperm is injected into the egg. After embryo transfer, extra embryos produced during the IVF cycle can be cryopreserved in a cryoprotectant and stored at low temperature until transfer during a later cycle. Couples who are at risk of transmitting a genetic disorder to their children can undergo PGD. Embryos are screened for genetic disease or age-related aneuploidy (for example, Down Syndrome) before embryo transfer. The technique involves the removal of one or two cells, or polar bodies, through a hole in the zona pellucida, to identify unaffected embryos by molecular analysis. Cytoplasmic transfer is a very uncommon procedure that has only been used in a few clinics. It is used for women who have repeatedly failed to become pregnant after IVF, and involves the transfer of cytoplasm, including mitochondria, from the eggs of young women to the eggs of older women.

One unresolved issue concerns the use of high doses of gonadotrophins to mature many eggs simultaneously. There is still doubt as to whether these drugs radically alter the uterine environment and the chances of successful implantation; studies in mice and hamsters suggest that superovulation decreases embryo and foetal viability^9,10. A number of human studies of the replacement of donated eggs and embryos, or freeze-thawed embryos, indicated that implantation may be more likely when large amounts of exogenous gonadotrophins are avoided. As we now have the molecular tools to analyse gene expression in the endometrium, it seems urgent that more studies of endometrial development and gene expression, with and without the use of these drugs, are undertaken. But the more serious effects of these drugs will probably be on the follicle and the oocyte maturing within. In animals and humans, many of the eggs aspirated after administration of these hormones are not viable. We do not know whether a fertilization rate of around 60% � the norm for an average IVF cycle � reflects what takes place normally in an unstimulated cycle inside the intact female genital tract. More importantly, studies of 'spare' human embryos collected after stimulated IVF cycles confirm a high incidence of developmental arrest, embryonic aneuploidy, mosaicism, apoptosis and failure of cytokinesis^11,12. Indeed, different superovulation regimens can alter both human pre-implantation embryo development and metabolism¹³. No more than 50% of IVF embryos complete pre-implantation development¹¹, and in general, fewer than 20% will implant satisfactorily and result in a live birth. Although these phenomena may be a result of unphysiologic culture, gonadotrophin administration may have a major influence. Interestingly, without IVF, singleton pregnancies after gonadotrophin treatment alone have a similarly increased risk of prematurity as IVF pregnancies, suggesting that the hormonal environment during oocyte maturation could be significant in neonatal outcome¹⁴. Finally, human embryos conceived both naturally and after IVF commonly have missing or extra chromosomes, and the majority of these chromosomal abnormalities may arise after resumption of meiosis, just before ovulation¹⁵. What is not clear is whether the incidence of chromosomal abnormalities is increased after administration of gonadotrophin, perhaps as a result of over-riding natural checkpoints. A major question of critical importance is what happens in natural conception cycles in humans. The ethical difficulties in obtaining naturally fertilized eggs seem so great that it may be very difficult to assess whether the 'mechanization' involved in routine IVF treatment induces anomalies in egg maturation. But it is just possible that high doses of gonadotrophins result in embryonic defects that are mostly lethal early in pre-implantation development, but not invariably.

Worldwide, a variety of culture media are used to culture embryos. Most media used in IVF clinics were originally based on classic recipes developed for the culture of cell monolayers or mouse embryos, whereas some more recent formulations have attempted to mimic the in vivo environment, despite the lack of detailed information about the composition of fluid in the fallopian tube. Commercial media purported to have been developed specifically for human embryo culture have increasingly been used in IVF centres, although their precise composition is not disclosed to the consumer. The high rate of developmental arrest during human pre-implantation development in vitro has resulted in a move towards extended culture, with the intention of selecting and transferring blastocysts that have survived the first five days in vitro¹⁶. However, the growing realization that the environment for gamete and pre-implantation development has far-reaching effects on subsequent foetal and postnatal development¹⁷ has caused unease about the safety of extended culture. Exposure of mouse embryos to different culture conditions alters expression^18,19 and imprinting^20,21 of a number of key genes, which could result in abnormal development. While many oocytes are aneuploid at the time of oocyte retrieval due to errors of meiosis, giving rise to aneuploid embryos¹⁵, chromosomal abnormalities can also arise after fertilization, during mitotic cleavage divisions^12,22. Despite the fact that these errors occur while the embryos are being cultured in vitro, there have been no studies investigating the impact of different culture media on chromosomal normality and the architecture of the mitotic spindle and cytoskeleton; this is an area of research that requires urgent attention.

Perhaps we should be more concerned about potential adverse effects of embryo cryopreservation. This is a common procedure in Western Australia and it is just possible that the effects that Hansen and colleagues reported are related to embryo freezing. Although early studies of human embryo freezing have not confirmed any teratogenic effects, nor an increased risk of producing foetal anomalies²³, the various studies to date have involved relatively small numbers of patients. It is a pity, therefore, that Hansen and her colleagues have not documented how many of their study patients allowed their embryos to be treated in such a manner, nor what cryoprotectants were used or at what stage of cleavage freezing was undertaken. There is a rational basis for being cautious about possible risks to the embryo after freezing. Embryo freezing is certainly associated with an attrition rate �after thawing, some embryos show loss of individual blastomeres or fail to undergo further development. Some cryoprotectants may also have deleterious effects on DNA²⁴. It is surprising that the potential of modern molecular approaches to detect mutations, such as transgenic mouse mutation assays, have not been exploited in an effort to investigate the possible mutagenicity of cryoprotectants and cryopreservation. It is well documented that different embryonic stages are differently affected by freezing, and optimal results are obtained by using an appropriate cryoprotectant for that stage. Moreover, equilibration of the cryoprotectant is important: too short a time, and there may be incomplete water molecule substitution; too long a time is known to increase the toxic effects of cryoprotectants. Furthermore, changes in embryonic metabolism during freezing may alter embryonic gene expression, especially in the later pre-implantation stages²⁵. There are also other possible molecular effects of cell freezing, telomere shortening and replicative senescence²⁶, damage to plasma and nuclear membranes, and inappropriate chromatin condensation. Moreover a subtle change in DNA might not manifest its effects until late in development, perhaps years after birth²⁷. For all these reasons, it is regrettable that more stringent follow-up is not routine after all IVF procedures of this type.

Another increasingly used IVF procedure is ICSI itself (Fig. 2). Most reports concerning ICSI were fairly encouraging²⁸, although a re-classification of cases from a large series suggested that the incidence of major defects had been under-estimated²⁹. This area needs careful re-evaluation. There are many other well-documented problems connected with this treatment, but one of the difficulties in assessing risk is that this procedure is quite often combined with others, such as cryopreservation or pre-implantation genetic diagnosis (PGD; Fig. 2). In a typical study of a mixed population, it was postulated that a slight increase in de novo chromosomal aberrations in children born after ICSI is more likely to be linked to the characteristics of the sperm injected rather than problems inherent in the procedure itself³⁰. However, this remains unproven. There is the obvious problem of injecting genetically abnormal sperm with consequences for health of the embryo: sperm which are themselves aneuploid or contain damaged DNA may cause anomalies of the foetus if injected into the egg³¹. But the procedure itself can result in potentially harmful substances being injected into the egg, leakage of cytoplasm, disruption of the meiotic spindle and membrane damage. It is known that embryos produced after ICSI have less potential to develop to the blastocyst stage than do eggs fertilized naturally³². Furthermore, the experience of the operator performing the procedure also makes a significant difference³³. Recently, rhesus monkey eggs, which show morphological similarities to human eggs, have been used as an animal model for ICSI. These studies have shown that abnormal nuclear remodelling results in asynchronous chromatin decondensation in the apical region of the sperm head, delaying the onset of DNA synthesis. The persistence of the acrosome and perinuclear theca on the apex of sperm introduced into the oocyte by ICSI may constrict the DNA in this region. Although normal monkeys have been born, it is worrying that the pressure of progress has meant that studies of this sort lag many years behind the onset of human treatments³⁴. However, the Belgian group that pioneered this technique has commendably collected a large database documenting neonatal outcome and congenital malformations after ICSI³⁵.

Although doubt remains over ICSI, even though ten of thousands of procedures have now been performed in Europe alone, there is much less experience with some other techniques that are being heavily promoted. Of these, the most important is embryo biopsy and PGD. At least with this procedure, extensive animal work in a number of species was performed before the first clinical treatments³⁶. However, 12 years after the first treatment, only a few hundred babies have been born worldwide. There have been no adverse effects reported in the offspring, but the chance of mis-diagnosis when analysing single blastomeres is high, and contamination at the time of gene amplification and allele dropout are two common problems. Moreover, the mosaicism of so many human embryos and the fact that apoptosis and other cell anomalies are common, means there is a risk that biopsy procedures may produce cells which do not reflect the whole embryonic genome. The removal of blastomeres is probably safe, as it is potentially less physically injurious than, say, ICSI or freezing. But whether this treatment should be extended to procedures such as aneuploidy screening in the hope that 'better quality' embryos may be transferred is still an open question, although clinics are now advertising, and charging for, this service.

A new therapy that has proceeded from concept to clinic with amazing rapidity is cytoplasmic transfer (Fig. 2). It has been proposed that oocytes from women who have suffered repeated IVF failure are of poor quality, with ageing mitochondria resulting in decreased energy production. Energy, in the form of ATP, is essential for accurate chromatid segregation at the time of fertilization and during subsequent mitotic divisions, as well as during blastocyst formation. It can therefore be envisaged that aged or defective mitochondria could result in aberrant chromosomal segregation or developmental arrest³⁷. It was thought that injection of a small amount of cytoplasm (containing mitochondria) from healthy donor oocytes would rejuvenate recipient oocytes from women with recurrent IVF failure³⁸. Over thirty children have been born worldwide through this technique of cytoplasmic transfer³⁹. However, the fact that the cells of the offspring will contain mitochondria from both donor and recipient oocytes (heteroplasmy) has raised significant concerns regarding the safety of this technique. Normally, mitochondria are thought to be uniform throughout the body, with a high degree of homoplasmy, which is thought to be important in the prevention of conflict between nuclear and mitochondrial DNA⁴⁰. Furthermore, mitochondria are maternally inherited, so if the child were female, both donor and recipient mitochondria could potentially be passed on to future generations. However, mitochondria are thought to pass through a bottleneck during oogenesis or embryogenesis, with clonal expansion from one or a few of these organelles⁴⁰. Thus, homoplasmy may be restored. Of significant concern was one study where two out of seventeen foetuses had an abnormal karyotype³⁹. Certainly, more experimental studies, particularly in animal models, are required before cytoplasmic transfer is routinely used in clinical practice.

All these worries are increased by the report from Schieve et al.⁴¹ that was published back-to-back with Hansen's data. In this report, they reviewed 42,463 IVF births and found that IVF babies were 2.6 times more likely to be underweight than those born after natural conception; this trend was significant, even after correction for such variables as multiple birth, maternal age and prematurity. Just possibly, embryo culture or embryo freezing could impair early development; given that Barker⁴² reports a clear relationship between low birth-weight and hypertension, cardiac disease, stroke and possibly osteoporosis in middle age, we must now be more vigorous in our surveillance of all these technologies.

It has been calculated that around one million babies have now been born as a result of IVF and related procedures, and the great majority are healthy. But there are legitimate concerns about the genetic risks, both potential and real, that are sometimes involved in this technology. Research with both animal and human embryos must continue, using modern molecular techniques to investigate the possible adverse effects of both cryopreservation and the culture environment on chromosomal and DNA integrity, gene expression and imprinting. It is highly desirable to consider how we can improve the collection and handling of data after these treatments, minimize the risks associated with assisted reproduction and regulate procedures in a more sensible manner. It is imperative that detailed and careful follow-up is carried out of children born as a result of assisted reproductive technologies, taking account of whether the embryos have been produced through classical IVF or by ICSI, cryopreserved, cultured for extended periods of time or had cells removed for PGD. Patient desperation, medical hubris and commercial pressures should not be allowed to be the key determining features in this generation of humans. Bringing a child into the world is the most serious human responsibility. We cannot ignore the clouds lowering over these valuable therapies. To do so could have a profound influence on the progress of medical science, not only in this high-profile field, but in others too.

References 1.  Hansen, M., Kurinczuk, J.J., Bower, C. & Webb, S.New Engl. J. Med.346, 725�730 (2002). 2.  MRC Working Party on Children Conceived by In Vitro Fertilisation. Br. Med. J.300, 1229�1233 (1990). 3.  French national registry, analysis of data 1986 to 1990. FIVNAT (French In Vitro National) Fertil. Steril.64, 746�756 (1995). 4.  Saunders, D.M. & Lancaster, P.Am. J. Perinatol.6, 252�257 (1989). 5.  Friedler, S., Mashiach, S. & Laufer, N.National Registry of the Israeli Association for Fertility Research. Hum. Reprod.7, 1159�1163 (1992). 6.  Doyle, P., Beral, V. & Maconochie, N.Hum. Reprod.7, 425�428 (1992). 7.  Verlaenen, H., Cammu, H., Derde, M.P. & Amy, J.J.Obstet. Gynecol.86, 906�910 (1995). 8.  Bergh, T.et al. Lancet354, 1579�1585 (1999). 9.  McKiernan, S.H. & Bavister, B.D.Hum. Reprod.13, 724�729 (1998). 10.  Van der Auwera, I. & D'Hooghe, T.Hum. Reprod.16, 1237�1243 (2001). 11.  Hardy, K.et al. Proc. Natl Acad. Sci. USA98, 1655�1660 (2001). 12.  Handyside, A. & Delhanty, J.Trends Genet.13, 270�275 (1997). 13.  Hardy, K.et al. Hum. Reprod.10, 2125�2135 (1995). 14.  Olivennes, F.et al. Hum. Reprod.8, 1297�300 (1993). 15.  Hassold, T. & Hunt, P.Nature Rev. Genet.2, 280�291 (2001). 16.  Gardner, D.et al. Fertil. Steril.69, 84�88 (1998). 17.  Kwong, W.Y.et al. Development127, 4195�4202 (2000). 18.  Ho, Y., Wigglesworth, K., Eppig, J.J. & Schultz, R.M.Mol. Reprod. Dev.41, 232�238 (1995). 19.  Niemann, H. & Wrenzycki, C.Theriogenology53, 21�34 (2000). 20.  Doherty, A.S.et al. Biol. Reprod.62, 1526�1535 (2000). 21.  Khosla, S., Dean, W., Reik, W. & Feil, R.Hum. Reprod. Update7, 419�427 (2001). 22.  Munn�, S., Weier, H.U., Grifo, J. & Cohen, J.Biol. Reprod.51, 373�379 (1994). 23.  Wada, I.et al. Hum. Reprod.9, 543�546 (1994). 24.  Bouquet, M., Selva, J. & Auroux, M.Biol. Reprod.49, 764�769 (1993). 25.  Emiliani, S.et al. Hum. Reprod.15, 905�910 (2000). 26.  Honda, S., Weigel, A., Hjelmeland, L.M. & Handa, J.T.Biochem. Biophys. Res. Commun.282, 493�498 (2001). 27.  Dulioust, E.et al. Proc. Natl Acad. Sci. USA92, 589�593 (1995). 28.  Van Steirteghem, A.et al. Hum. Reprod.13, S134�S142 (1998). 29.  Kurinczuk, J.J. & Bower, C.Br. Med. J.315, 1260�1266 (1997). 30.  Bonduelle, M.et al. Hum. Reprod.11, S131�S159 (1996). 31.  Mroz, K., Hassold, T.J. & Hunt, P.A.Hum. Reprod.14, 1151�1156 (1999). 32.  Dumoulin, J.C.et al. Hum. Reprod.15, 402�409 (2000). 33.  Dumoulin, J.M.et al. Hum. Reprod.16, 306�312 (2001). 34.  Hewitson, L., Simerly, C., Dominko, T. & Schatten, G.Theriogenology53, 95�104 (2000). 35.  Bonduelle, M.et al. Hum. Reprod.17, 671�694 (2002). 36.  Handyside, A.H., Kontogianni, E.H., Hardy, K. & Winston, R.M.Nature344, 768�770 (1990). 37.  Van Blerkom, J., Sinclair, J. & Davis, P.Hum. Reprod.13, 2857�2868 (1998). 38.  Cohen, J.et al. Mol. Hum. Reprod.4, 269�280 (1998). 39.  Barritt, J., Willadsen, S., Brenner, C. & Cohen, J.Hum. Reprod. Update7, 428�435 (2001). 40.  Cummins, J.M.Hum. Reprod. Update7, 217�228 (2001). 41.  Schieve, L.A.et al. N. Engl. J. Med.346, 731�737 (2002). 42.  Barker, D.J.Proc. R. Soc. Lond. B262, 37�43 (1994).